1. Field of the Invention
The present invention relates to a video encoding apparatus which is applied, for example, to a video communication system.
2. Description of the Prior Art
FIG. 1 shows an example of such apparatus in the prior art, particularly a video encoding apparatus of the movement compensation type. In FIG. 1, numeral 1 designates an A/D converter which receives video signals outputted from a video camera 8 and converts the signals in analog/digital conversion and then outputs them. A block constituting member 2 receives the digital signal series outputted from the A/D converter 1 and makes every K picture elements close on the picture image (K: integer not less than 2) in a block and outputs the signals in the block. A frame memory 3 stores data transmitted one frame before the existing input signal frame and supplied through an adder 10. A reference signal generator 4 reads from the frame memory 3 the input block including the existing input signal series constituted into a block in the block constituting member 2 and also plural blocks including the block at the same position on the picture image and formed by the signal series which is one frame before the existing frame. A distortion operation member 5 operates distortion between the signal series of the input block outputted from the block constituting member 2 and the signal series of the plural blocks stored in the frame memory 3, and outputs the position information of the block giving the minimum distortion and the existing signal series. A subtractor 6 estimates difference signal series (vector) between the signal series of the input block outputted from the block constituting member 2 and the signal series of the block giving the minimum distortion and outputted from the distortion operation member 5, and then outputs the difference signal. A quantization encoding member 7 receives the difference signal series (vector) outputted from the subtractor 6 and performs quantization encoding of the difference signal. A quantization decoding member 9 reproduces the difference signal series from the quantization encoding signal outputted by the quantization encoding member 7. An adder 10 adds the quantization decoding output from the quantization decoding member 9 to the signal series of the block giving the minimum distortion and outputted from the distortion operation member 5, and reproduces the video signal and writes it to the frame memory 3.
The principle of the vector quantization will now be briefly described referring to FIG. 2.
The information source input signal series of K in number are brought together into input vector x={x.sub.1, x.sub.2, . . . , x.sub.K }. Then, the K-dimensional Euclidean signal space R.sup.K (x.epsilon.R.sup.K) has the representative points of N in number (i.e., output vector) y.sub.i ={y.sub.i1, y.sub.i2, . . . , y.sub.iK }, and set of the representative points is made Y={y.sub.1, y.sub.2, . . . , y.sub.n) . If each partition of having output vector y.sub.i as the representative point (e.g., center of gravity) is made R.sub.1, R.sub.2, . . . , R.sub.N, vector quantization Q is defined by following formula: ##EQU1## The vector quantization Q is expressed as cascade connection of encoding C and decoding D. The encoding C is mapping of the output vector set Y={y.sub.1, y.sub.2, . . . , y.sub.N } of R.sup.K to index set I={1, 2, . . . , N}, and the decoding D is mapping from I to Y. That is, EQU C:R.sup.K .fwdarw.I, D:I.fwdarw.Y (4) EQU Q=D.multidot.C (5)
Since the encoding output I is transmitted or recorded in the vector quantization, the encoding efficiency is quite good.
The vector quantization is the mapping of the input vector to the output vector y.sub.i in the minimum distance (minimum distortion). More specifically, if distance (distortion) between input/output vector is made d(x, y.sub.i), it follows that EQU if d(x, y.sub.i)&lt;d(x, y.sub.i) for all j (6) EQU x.epsilon.R.sub.i hence x.fwdarw.y.sub.i ( 7)
Set Y of the output vector y.sub.i as shown in FIGS. 3(A) and 3(B) can be determined by clustering using the information source input signal series being the training model (repetition of selection of the representative points and the partition of signal space until the total distortion becoming minimum).
Operation of the apparatus in above-mentioned constitution of FIG. 1 will be described. Video signals outputted from the video camera 8 to the A/D converter 1 are converted into digital signal series by the A/D converter 1 and then outputted to the block constituting member 2. In the block constituting member 2, every K picture elements close on the picture image are made a block usually in rectangular form (or square form) on the picture image, and transformed in the arrangement and then outputted to the distortion operation member 5 and the subtractor 6. On the other hand, the signal series of the plural blocks, which are stored in the memory 3 and read by the reference signal generator 4, are transmitted to the distortion operation member 5 in synchronization with the existing input signal series outputted from the block constituting member 2 by the reference signal generator 4. An example of the position relation between the block of the existing input signal series outputted from the block constituting member 2 and the plural blocks read by the reference signal generator 4 is shown in FIG. 3(A) illustrating the existing frame and FIG. 3(B) illustrating the frame being one before the existing frame. The distortion calculation between the signal series of the existing input block and the signal series of the plural blocks is performed in the distortion operation member 5 using, for example, Euclidean distortion or absolute value distortion. According to the calculation results, the block to provide the minimum distortion to the existing input block is selected among the plural blocks. Assuming that the blocks as the calculation object are M in number (M: integer not less than 2) and the block to provide the minimum distortion is the l-th one among the M blocks (l=1, 2, . . . , M), the distortion operation member 5 outputs the value of l and the signal series of the block to the subtractor 6. The subtractor 6 estimates the difference signal series (vector) between the signal series of the input block and the signal series to give the minimum distortion and outputs the difference signal series to the quantization encoding member 7. The difference signal series (vector) is subjected to the quantization encoding in the quantization encoding member 7 and then outputted to the quantization decoding member 9. Both the difference signal series reproduced by the quantization decoding member 9 and the signal series of the block outputted from the distortion operation member 5 and giving the minimum distortion are outputted to the adder 10 and added by the adder 10 an reproduced in the video signal. The video signal is written in the frame memory 3 by the adder 10. The signal processing in the process from the quantization encoding member 7 up to the adder 10 in effect relates to the subsequent processing by vector of the difference signal series estimated in the subtractor 6. The smaller the amount of the difference signal, i.e., the vector quantity, the higher the efficiency of the movement compensation. In the amount as described above, the inputted video signal is formed as a block which is used to select the block most resembling those in the former frame. The newest block is compared to the selected block, the difference or error component being quantized to form a coded output signal, together with information representing which block is the most resembling one.
The video encoding apparatus in the prior art, particularly the video encoding apparatus of the movement compensation type is constituted as above described. For example, even in the movement such as the oscillating motion of the video camera 8 where the moving direction of the screen is clear, the efficiency of the movement compensation cannot be increased in the prior art, because the position relation between the block as object of the distortion operation stored in the frame memory 3 and the input block outputted from the block constituting member 2 is fixed.